Statistical methods for surveillance of congenital malformations: when do the data indicate a true shift in the risk that an infant is affected by some type of malformation?

Abstract

Statistical methods for surveillance are reviewed for use with both hospital-based systems and central registries. In the hospital-based system the surveillance methods are applied as each new case occurs and the methods focus on the number of unaffected births between each case. In centralized systems, it is usually more convenient to observe the number of cases in time intervals of fixed length. Methods for calculating exact confidence limits about the number of cases, the proportion of malformed cases and the observed-to-expected ratio are reviewed, as are methods allowing evidence to accumulate over several time periods. Examples are given to illustrate the use of the different methods.